Manual correction of conventional exam answer sheets is a primary problem due to being time-consuming and prone to errors, while commercial OMR devices are costly. This research aims to design and implement an efficient and accurate computer vision-based Optical Mark Recognition (OMR) system for conventional answer sheets. The method employed is an image processing pipeline using OpenCV, which includes preprocessing with Adaptive Thresholding, contour detection, perspective transformation for skew correction, and answer area segmentation for choice extraction based on pixel analysis. Testing results on 30 samples for each condition showed that the system achieved 100% accuracy on thick markings, 99.22% on thin markings, and 86.22% on thin markings with significant scribbles. It is concluded that the developed system is highly effective for answer sheets with clean markings, but its performance degrades when visual noise from scribbles is present, thus offering an affordable OMR alternative with identified performance limitations.